Measuring neural activity in the brain is useful for medical diagnostics, neuromodulation therapies, neuroengineering, or brain-computer interfacing. For example, it may be desirable to measure neural activity in the brain of a patient to determine if a particular region of the brain has been impacted by reduced blood irrigation, a hemorrhage, any other type of damage. For instance, in cases where the patient has suffered a traumatic brain injury, such as stroke, it may be desirable to determine whether the patient should undergo a therapeutic procedure. Measuring neural activity in the brain also may be used to determine the efficacy of such a therapeutic procedure.
Conventional methods for measuring neural activity in the brain include diffuse optical tomography (DOT), and functional near-infrared spectroscopy (fNIRS), as well as others. These applications only employ a moderate amount of near-infrared or visible light radiation, thus being comparatively safe and gentle for a biological subject in comparison to X-Ray Computed Tomography (CT) scans, positron emission tomography (PET), or other methods that use higher-energy and potentially harmful radiation. Moreover, in contrast to methods, such as functional magnetic resonance imaging (fMRI), these optically-based imaging methods do not require large magnets or magnetic shielding, and thus, can be scaled to wearable or portable form factors, which is especially important in applications such as brain-computer interfacing.
Because DOT and fNIRS rely on light, which scatters many times inside brain, skull, dura, pia, and skin tissues, the light paths occurring in these techniques comprise random or “diffusive” walks, and therefore, only limited spatial resolution can be obtained by a conventional optical detector, often on the order of centimeters. The reason for this limited spatial resolution is that the paths of photons striking the detector in such schemes are highly variable and difficult, and even impossible to predict without detailed microscopic knowledge of the scattering characteristics of the brain volume of interest, which is typically unavailable in practice (i.e., in the setting of non-invasive measurements through skull for brain imaging and brain interfacing). In summary, light scattering prevents optical imaging from achieving high resolution deep inside tissue.
There is an increasing interest in ultrasound modulated optical tomography (UOT) to detect more precisely localized changes in biological tissues, e.g., on a sub-millimeter length scale, inside thick biological tissue, such as the brain (see U.S. Pat. No. 8,423,116; Sakadzic S, Wang L V, “High-Resolution Ultrasound-Modulated Optical Tomography in Biological Tissues,” Optics Letters, Vol. 29, No. 23, pp. 2770-2772, Dec. 1, 2004). These localized changes may include changes in light absorption in the brain that reflect neural activity and neurovascular coupling, such as a blood-oxygen-level dependent signal, for application in diagnostics, therapeutics, or, notably, brain computer interfacing (see Steinbrink J, Villringer A, Kempf F, Haux D. Boden S, Obrig H., “Illuminating the BOLD Signal: Combined fMRI-fNIRS Studies,” Magnetic Resonance Imaging, Vol. 24, No. 4, pp. 495-505, May 31, 2006). Thus, there is an increasing interest in ultrasound modulated optical tomography (UOT) in biomedical applications due to its potential to simultaneously achieve good resolution and imaging depth.
In UOT, a highly localized ultrasound focus, e.g., millimeter or sub-millimeter in size, is used to selectively perturb (i.e., “tag”) light (e.g., light generated by a near-infrared coherent laser) passing through a voxel size of tissue defined by the size of the ultrasound focus. Due to the acousto-optic effect, light passing through the ultrasonic beam undergoes a frequency shift defined by multiples of the ultrasonic frequency. By detecting the frequency-shifted light, i.e., the tagged light, spatial information characterizing the biological tissue within the voxel can be acquired. As a result, spatial resolution is boosted from the centimeter-scale diffusive spread of light in the biological tissue to approximately a millimeter-scale voxel size. This ultrasound tagging of light relies on mechanisms known in the field (see Mahan G D, Engler W E, Tiemann J J, Uzgiris E, “Ultrasonic Tagging of Light: Theory,” Proceedings of the National Academy of Sciences, Vol. 95, No. 24, pp. 14015-14019, Nov. 24, 1998).
Typical UOT implementations generate weak signals that make it difficult to differentiate ultrasound-tagged light passing through the focal voxel from a much larger amount of unmodulated light which is measured as DC shot noise. Thus, conventional UOT has the challenge of obtaining optical information through several centimeters of biological tissue, for example, noninvasive measurements through the human skull used to measure functional changes in the brain.
Various methods have been developed to detect the very small fraction of tagged light out of a large background of untagged light by detecting the speckle pattern of light resulting from the interference of many multiply-scattered optical waves with different phases and amplitudes, which combine in a resultant wave whose amplitude, and therefore intensity, as well as phase, varies randomly. In the context of neuroengineering and brain computer interfacing, a key challenge is to render these methods to be sufficiently sensitive to be useful for through-human-skull functional neuroimaging.
One technique uses a narrow spectral filter to separate out the untagged light striking a single-pixel detector, and is immune to speckle decorrelation (greater than ˜0.1 ms-1 ms) due to the scatters' motion (for example, blood flow) inside living biological tissue, but requires bulky and expensive equipment.
Another technique uses crystal-based holography to combine a reference light beam and the sample light beam into a constructive interference pattern, but can be adversely affected by rapid speckle decorrelation, since the response time of the crystal is usually much longer than the speckle correlation time.
Still another technique, referred to as heterodyne parallel speckle detection (PSD), employs optical interference together with a spatially resolved detector array (e.g., a conventional charge-coupled device (CCD) camera) used as an array of independent detectors for collecting the signal over a large number of coherence areas (see Atlan M, Forget B C, Ramaz F, Boccara A C, Gross M, “Pulsed Acousto-Optic Imaging in Dynamic Scattering Media With Heterodyne Parallel Speckle Detection,” Optics Letter, Vol. 30, No. 11, pp. 1360-1362, Jun. 1, 2005). Such configuration improves the signal-to-noise ratio relative to a single-detector and relative to approaches based on other modes of separating tagged and untagged light, such as spectral filters. However, the conventional CCD cameras used for heterodyne PSD have low frame rates, and therefore suffer from a relatively low speed relative to the speckle decorrelation time, thereby making this set up insufficient for in vivo deep tissue applications. Furthermore, conventional CCD cameras record both the AC signal and the DC background for each pixel. Thus, only a few bits of a pixel value can be used to represent the useful AC signal, while most of the bits are wasted in representing the DC background, resulting in a low efficiency in the use of bits.
Lock-in cameras, as compared to conventional CCD cameras, have been used for comparatively bit-efficient and noise resistant heterodyne PSD (see Liu Y, Shen Y, Ma C, Shi J, Wang L V, “Lock-in Camera Based Heterodyne Holography for Ultrasound-Modulated Optical Tomography Inside Dynamic Scattering Media,” Applied Physics Letters, Vol. 108, No. 23, 231106, Jun. 6, 2016; see also Liu Y, Ma C, Shen Y, Wang L V, “Bit-Efficient, Sub-Millisecond Wavefront Measurement Using a Lock-In Camera for Time-Reversal Based Optical Focusing Inside Scattering Media,” Optics Letters, Vol. 41, No. 7, pp. 1321-1324, Apr. 1, 2016). For each pixel, a lock-in camera is capable of performing lock-in detection and outputting only information of the AC signal as a single AC amplitude map that is transferred to a computer, and thus, provides an efficient means of detecting and processing the speckle pattern.
Besides the challenges posed by the signal-to-noise ratio, speckle decorrelation time, and efficient pixel bit processing, another challenge involves obtaining sufficient axial resolution (i.e., the depth resolution or ultrasound propagation direction). To address this challenge, UOT has been applied in a pulsed wave (PW) mode for heterodyne PSD, rather than a continuous (CW) mode (see Li Y Zhang H, Kim C, Wagner K H, Hemmer P., Wang L V, “Pulsed Ultrasound-Modulated Optical Tomography Using Spectral-Hole Burning as a Narrowband Spectral Filter,” Applied Physics Letters, Vol. 93, No. 1, 011111, Jul. 7, 2008; Ruan H, Mather M L, Morgan S P, “Pulsed Ultrasound Modulated Optical Tomography with Harmonic Lock-In Holography Detection,” JOSA A, Vol. 30, No. 7, pp. 1409-1416, Jul. 1, 2013).
PW UOT has the benefit of enabling improved axial resolution compared to CW UOT. That is, with CW UOT, any light passing through the tissue, even though outside of the focal voxel, may be inadvertently tagged by the continuously propagating ultrasound energy along the ultrasound axis, thereby decreasing the signal-to-noise ratio. With PW UOT, the light passing through the tissue is pulsed only when the ultrasound pulses travels through the focal voxel, such that light outside of the focal voxel will not be tagged by the ultrasound energy. Although PW UOT improves axial resolution, the pulsed UOT signals are weak relative to continuous UOT signals.
Although the UOT schemes described above may be sufficient for certain applications, such UOT schemes are inappropriate for the application of 3D-resolved, highly sensitive detection of small signals (e.g., blood-oxygen-level dependent signals) non-invasively through thick scattering layers, such as the human skull.